Generative Artificial Intelligence
Wayfinding: Tools > Other Tools > Generative Artificial Intelligence
A thorough discussion of the benefits and detriments of the current large language models and image generators in the most recent versions of generative artificial intelligence (AI) is beyond the scope of this project; it is worth noting that there are several violations of intellectual property and copyright under investigation due to how these tools are trained, there are jobs in creative industries actively being lost to AI, AI is generally found to provide inaccurate responses when asked questions that would otherwise be directed to search engines, it is getting worse over time and not better, and the computational power required to use AI is already destroying ecosystems. That said, in 2024, artificial intelligence has been thrust into the public consciousness and with massive international corporations like Meta, Microsoft, and Apple incorporating AI tools into their products, it seems that creative workers have no choice but to test them out. Like other digital tools, though, AI has been integrated into theatre performances since before the pandemic; it just seems especially dire now. Susan Broadhurst is perhaps the originator of AI agents onstage when she created a performance in 2001 called Blue Bloodshot Flowers ‘which explores and examines this physical/virtual interface from the audience but also from the performer’s perspective’. Elodie, a dancer and the main character, performs a bittersweet memory of a lost lover; her co-star, Jeremiah, is a large, disembodied head powered by an early artificial intelligence. Jeremiah could ‘express himself and his emotions, such as, anger, sadness or happiness. … he can demonstrate several emotions as a reaction to visual stimulus’. This was due to front-facing sensors or ‘eyes’ which allowed the AI to experience the dance performance onstage, as well as facially react to audience members who engaged him. As a method of proving that Jeremiah was not pre-recorded and was reacting in real-time, Broadhurst describes:
At the 291 Gallery, audience members arrived right up until the very end of the scripted performance since we had decided not to restrict entrance. We had allowed unrestricted entrance for the very reason that Jeremiah would interact with any new arrivals he spotted and of course he did, which amused everyone except possibly the late arrivals.
Broadhurst also notes that Jeremiah routinely displayed ‘inappropriate’ or ‘chaotic’ behaviour and implies this goes with his programming that he was sad or bored when left alone. Broadhurst’s pioneering work with the help of Surrey University’s Richard Bowden, a systems engineer and AI programmer, demonstrates some of the potential behavioural issues AI still displays, which in extremis are hallucinations or catastrophic forgetfulness, leading to inappropriate responses in our current testing. Though Jeremiah remained ‘in character’, his stage presence was largely as a reactionary agent – he was not able to speak aloud, only facially react, and this did not change the unfolding of the memory dance, though it likely had a huge impact on the audience’s interpretation of the narrative.
Broadway is not shying away from integrating A.I. creations onto stage, seeming to pick up where Broadhurst left off. Robert Downey Jr. makes his Broadway debut alongside a 'metahuman' recreation of him in McNeal by Ayad Akhtar in 2024. Seemingly an update on Samuel Beckett's Krapp's Last Tape in which a man obsessed with media and himself listens to recordings of his past experiences, McNeal follows a writer fighting three foes: 'an estranged son, a new novel, old axes to grind and an unhealthy fascination with Artificial Intelligence' which includes a full digital recreation of himself, onstage and speaking to him. The show, in part, aimed to create a legal, authorised, within-copyright version of the popular film actor which could potentially help them differentiate 'any non-consensual fakes that crop up'. Over the course of this research, I was fortunate to watch Brendan Bradley, Clemence Debaig, and Michael Morran attempt to engage image generators in an ethical way, using them to produce assets for their shows in OnBoardXR 6: it’s pronounced ‘gybe’.
Last updated